Real-time 3D Skeletonisation in Computer Vision-Based Human Pose Estimation Using GPGPU

Human pose estimation is the process of approximating the configuration of the body's underlying skeletal articulation in one or more frames. The curve-skeleton of an object is a line-like representation that preserves topology and geometrical information. Finding the curve-skeleton of a volume corresponding to the person is a good starting point for approximating the underlying skeletal structure. In this paper a GPU implementation of a fully parallel thinning algorithm based on the critical kernels framework is presented. The algorithm is compared to another state-of-the-art thinning method, and while it is demonstrated that both achieve real-time frame rates, the proposed algorithm yields superior accuracy and robustness when used in a pose estimation context. The GPU implementation is > 8 x faster than a sequential version, and the positions of the four extremities are estimated with rms error similar to 6 cm and similar to 98 % of frames correctly labelled.